Worked example · Project management

How I run projects with AI.

One project, from first ask to delivery. The same project management I've done for years, rebuilt around AI. It's faster to set up, and it keeps up when things change.

Project Consumer research and brand strategyClient A light beer brand (fictional)Length 12 weeksPlanning Up to five days before. Under half that now.

About this example

The example is fictional. It's based on the shape of real research projects I've run, but the client, market, people and figures are made up. A light beer brand that used to lead its category is losing ground to a newer rival, and the client needs to know why. What follows is how that project gets planned, tested and run.

The plan and the work Where I decide Results coming back

0 The old way

Five days of admin before the real work started.

Every project began the same way. A set-up form. A timeline built cell by cell in a spreadsheet. A Gantt chart. A status tracker. On a big project, that was up to five days. Then the first thing changed, and half of it needed redoing by hand.

1 The starting point

It starts with what the client asked for.

A light beer brand that used to lead its category is losing ground to a newer rival. The client has the sales figures. What they're missing is the why. They want research, a brand refresh, more awareness and a return to growth, all in 12 weeks.

2 Sharpen the job

Find the problems before anyone spends a cent.

Before a plan exists, I test the ask against foundational marketing theory. What is it really relying on? Do the objectives and measures agree? Is there a real insight, or just a description of the audience? This one was trying to do two projects at once and be five things at once.

AI does

Reads the request line by line. Flags contradictions, gaps and weak insight.

I do

Decide which flags matter, and what the project commits to.

3 The team

The job decides who's needed.

With the job clear, the roles follow. A research project needs recruitment, a researcher in the room, someone to run the workshops, and a strategist to pull it together. It doesn't yet need a creative team. That comes after.

4 The plan

Twelve weeks, planned in half the time.

The timeline, the Gantt chart, the status tracker and the pre-read schedule, all drafted from the job and checked by me. Every sign-off point is set from the start. Pre-reads go out three days before each workshop. I learned that one the hard way.

5 Delivery

When the plan changes, the documents change with it.

In week 4, focus group recruitment slipped by two weeks. In the old way, that meant a day of rebuilding spreadsheets. Now the plan works out the knock-on effects straight away and sets out the options. It can't decide what the client is told. That part is mine.

AI does

Works out what the slip touches, sets out the options, updates every document and drafts the note to the client.

I do

Choose the option, and approve what the client hears. Here: move the debrief. Rushing the analysis would weaken the one thing the client is paying for, the why.

Less time on the admin.

More time on the work.

Illustrative. Based on the shape of a real project, with the client, market and figures changed.